[
https://issues.apache.org/jira/browse/CASSANDRA-10140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14705573#comment-14705573
]
Ariel Weisberg edited comment on CASSANDRA-10140 at 8/20/15 8:38 PM:
---------------------------------------------------------------------
Trying to establish the impact empirically is not going to be easy. If you turn
on gc logging and run cstar I would not expect to see an impact.
This something that happens pretty rarely even on configurations that suffer
from it. You really do have to get unlucky. So maybe it doesn't matter. Pauses
occur out the yin yang anyways so the benefit of having GC logging will easily
outweigh the occasional hiccup. It's a barn door scenario if you are doing
mmapped reads since those will randomly extend TTSP.
I am just socializing the concern not -1ing it.
We did go to the trouble of disabling one other source of this kind of
infrequent [JVM writes to the filesystem blocking the VM
already.|https://issues.apache.org/jira/browse/CASSANDRA-9242] This would be a
step in the opposite direction IF the JVM doesn't do the logging correctly.
was (Author: aweisberg):
Trying to establish the impact empirically is not going to be easy. If you turn
on gc logging and run cstar I would not expect to see an impact.
This something that happens pretty rarely even on configurations that suffer
from it. You really do have to get unlucky. So maybe it doesn't matter. Pauses
occur out the yin yang anyways so the benefit of having GC logging will easily
outweigh the occasional hiccup.
I am just socializing the concern not -1ing it.
We did go to the trouble of disabling one other source of this kind of
infrequent [JVM writes to the filesystem blocking the VM
already.|https://issues.apache.org/jira/browse/CASSANDRA-9242] This would be a
step in the opposite direction IF the JVM doesn't do the logging correctly.
> Enable GC logging by default
> ----------------------------
>
> Key: CASSANDRA-10140
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10140
> Project: Cassandra
> Issue Type: Improvement
> Components: Config
> Reporter: Chris Lohfink
> Assignee: Chris Lohfink
> Priority: Minor
> Attachments: CASSANDRA-10140.txt
>
>
> Overhead for the gc logging is very small (with cycling logs in 7+) and it
> provides a ton of useful information. This will open up more for C*
> diagnostic tools to provide feedback as well without requiring restarts.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
